Transcript DNDO - All Slides For Public Release

Position Sensitive Dual-Layer Silicon Diode
Micro-Array Detector
Andrew Taylor, Yoonseok Yang, Gwan Choi
Dept. of Electrical and Computer Engineering, Texas A&M University
Position Sensitive Dual-Layer Silicon Diode Micro Arrays Detector
•Can determine the incident angle at the resolution of 0.006 degrees with dual-layer position
sensitive detector arrays
•Detection time resolution can be set to 1nS or faster
•Can estimate the energy level of the particle
•No expensive/special
fabrication processes
(cheapest 2-metal layer
CMOS processing)
•Wireless communication
Using latest
“cooperative modulation”
•Reconfigurable and
redundant design to
filter manufacturing
defects w/o testing
Position Sensitive Dual-Layer Silicon Diode Micro Arrays Detector
•Over 91% active area for efficiency
•Novel DSP to handle over 250 quintillion pixels per second (compared to 120+Giga pixels per
second for State-of-the-art Canon DIGIC4 CMOS High-Definition Image Sensor)
•Passive sensing for Ultra-Low Power operation
•Signal processing and transmission doesn’t start until detection event
•Diode current leakage is the only dominant source standby power
•Battery life orders of magnitude lower than
that of active sensors
•Array is dense (1M detection elements per 1mm2)
•Reconfigurable NOC will support manufacturing
of wafer-scale detector with high yield
•e.g. ~approximately $20-25(today) for
8 inch wafer detection
•Total cost approximately $110 for mass produced
2 x 124.6billion – element detector
Figure 2. Array detector Architecture
Photodiode array design
This array design has similarities to a CMOS image sensor array
The traditional CMOS image sensor precharges
a photodiode and reads the output every clock cycle.
This requires overhead transistors as seen in the
diagram below. Although the CMOS image sensor can
be used for the position sensitive detection of nuclear
particles, they constantly consume power and are
best left to be used in cameras
We expect particles to strike the array at
random locations and random times. Rather
than read signals synchronously like the CMOS
image sensor, an asynchronous passive-mode
solution is ideal for this application. A current
spike flows through one row and one column of
our array design below when a photodiode is
stuck by radiation.
1x1 micron element array has over 91% visible active area
Layout Illustraion
(Not to scale)
Metal layer 2
Metal layer 1
Polysilicon
N-P active area
Traditional charge-particle model is used to simulate photodiode current caused by a
radiation strike. LET (Linear Energy Transfer) values quantify the extent of radiation
strike. We studied LET values from 1 to 20, 1 being a weak strike and 20 being a strong
strike. The graph below shows that we have a 14X difference between the magnitude
of a LET1 and LET20 current. Due to this wide range of current that we want to detect,
a novel differential pairs are developed to amplify the signal.
•The amplifiers connected to row and column to
amplify voltage across a resistor to a bias source
The circuit was simulated with SPICE
•Results show that LET1 currents (weakest) are
detectable by an output inverter for both the NMOS
and PMOS differential pairs
•Output inverters were sized to sit at the threshold
and trigger with a small voltage as shown in the
graphs
Input and Output Voltage Response of PMOS Differential Pair
Row Amplifier with Weak Voltage Spike
Input and Output Voltage Response of NMOS Differential
Pair Row Amplifier with Weak Voltage Spike
Circuit simulations are conducted to simulate LET1 current spikes with varying photodiode
sizes (60nm – 1um width) and varying array sizes. From these batch simulations, we were
able to determine design constraints for the array. The graph below shows maximum
array size for each photodiode width.
Figure 3. Large-area bidirectional serial link for array-detector architecture
Detection signal processing : Draft 1. Monitor each pixel??
Detection signal processing : Draft 2.
 Data compression before sending it to main controller
Run length coding or Huffman coding?
 Data decompression
Main controller has a few decompression blocks and
share them to decompress received data
Pros: HW is simple and robust for detection
Architecture(Block and MB size) is reconfigurable
Compression of data is done on stable condition
Cons: Lossy approach (1 detection on a MB)
It still needs a bunch of signals (Ex.1000 signals)
Detection signal processing : Draft 2.
Detection signal processing : Draft 2.
Detection signal processing : Draft 3.
Detection signal processing : Draft 3.
Detection signal processing : Draft 4.
Initial block logic diagram
Address generation
Resolution generation
Detection process
Detection signal processing : Draft 5.
Detection Simulation Result
Detection Simulation Result
Detection Simulation Result
Design Synthesis Result
****************************************
Report : area
Design : rd_top
Version: C-2009.06-SP2
Date : Mon Nov 23 12:45:59 2009
****************************************
Library(s) Used:
NangateOpenCellLibrary_PDKv1.2_v2008_05 (File: /../../project/LIB/Nangate_typical_ccs.db)
Number of ports:
1041
Number of nets:
34451
Number of cells:
31249
Number of references:
55
Combinational area:
26513.017812
Noncombinational area: 14106.512474
Net Interconnect area:
undefined (No wire load specified)
Total cell area:
Total area:
40619.530286
undefined
Position Sensitive Dual-Layer Silicon Diode Micro Arrays Detector
S. “Gwan” Choi [email protected], Associate Professor
Electrical & Computer Engineering Department
Texas A&M University
ph)979-820-2553
•B.S., M.S., Ph.D. all in Electrical and Computer
Engineering, all from University of Illinois at Urbana
•Radiation susceptibility analysis and design experiences
over 20 years
•NSF CAREER award recipient
•Taught and conducted CMOS VLSI research at Texas A&M
for the past 15 years
•Patent for a low-power communication architecture
Key Claims of the presented microarray detector
•Can determine the incident angle at the resolution of 0.006
degrees with dual-layer detector arrays
•Can estimate the energy level of the particle
•Detection time resolution can be set to 1nS or faster
•No expensive/special fabrication steps (cheapest 2-metal
layer CMOS processing) to manufacture these detectors
•Wireless communication using latest “cooperative
modulation” using MIMO technology
•Reconfigurable and redundant design to filter
manufacturing defects w/o testing
Attributes of the detector design
•Over 91% active area for efficiency
•Novel DSP to handle over 250 quintillion pixels per
second
•Passive sensing for Ultra-Low Power operation
•Array is dense (1M detection elements per 1mm2)
•Reconfigurable NOC will support manufacturing
of wafer-scale detector with high yield
e.g. ~approximately $20-25(today) for
8 inch wafer detection
•Total cost approximately $110 for mass produced
2 x 124.6billion – element detectors